Thought I'd kick off this topic with a reference to this weeks reading on the Semantic Web, in Scientific American.
I found the software agents diagram particularly interesting as it points to the potential of agents in the front and back end using the notion of a semantic web.
One thought I had while reading the article is whether the idea of a error free semantic web can really exist. The article is convincing, but after considering the idea of "knowledge brokers" in the first week of class, I wonder if there might be a faster, more efficient semantic web? That is, one agent serves the front end, the semantic web provides for a backend (system to system) agent AND - there is a knowledge broker agent which serves up the information from the backend agent more efficiently by aggregating information. Just a thought - might be making it more complicated than it needs to be.
Any other thoughts?
Amanda
Posted by Amanda at February 10, 2003 09:05 PMThe best kind of agent would be one that is either verified for accuracy/quality by an actual person, or at worst - that several people use (a vote by use) that can fit into a model put into a system. The semantic represenation of data could help with this by forcing agents to act set ways depending on the type of data and its context.
Yes, Collaborative Filtering is one way to do that. The challenge is to make this be efficient in terms of the "knowledge market" concept.
It appears on a closer look, many of our class topics are starting to blur together...
Posted by: donturn on February 10, 2003 10:35 PM